Money Basics

Backtest Truth, Part 2: How Fees Turn a Winner Into a Loser

Fees of 0.07% and 0.10% per side changed the same rules from about +245% to about +178% in 2 years, and from about +77% to about +8% on older data.

On this page
  1. What “per side” means
  2. A strong-looking setup, then a higher fee
  3. Fewer trades, smaller damage
  4. Rules that entered very often
  5. What an honest fee test looks like
  6. Key takeaways
  7. FAQ

This is education only and not financial advice. Trading involves risk, and you can lose money.

Part 1 showed that a test can invent profit by filling at a price you could not have got. Part 2 is quieter. Every time a trade opens or closes, a cost comes out. The rules can stay clever. The fee does not care.

What “per side” means

In these tests the base cost was 0.07% per side, and 0.10% per side was a stricter check. “Per side” means you pay once to enter and once to exit. A round trip at 0.07% costs 0.14%. At 0.10% it costs 0.20%.

On one trade that looks tiny. It is not tiny when you trade often, or when the edge is already small. An edge is the small advantage the rules hope to have after everything honest is counted. If the hope is thinner than the bill, the test was never a plan.

Think of a remittance fee. One small charge is easy to ignore. The same charge on every transfer, all year, is rent. Trading fees work the same way. They come out first, before the story about being clever.

A strong-looking setup, then a higher fee

One of the stronger setups saved in this work was a 1-hour pullback in the direction of a 4-hour and weekly filter. At 0.07% per side it looked strong: about +90% in 1 year and +245% in 2 years. At 0.10% per side those results fell to about +69% and +178%.

The rules did not change. The cost of trading did. A move from 0.07% to 0.10% per side is a small step on a fee schedule. On this setup it took a large bite out of both windows. The 2-year figure fell from about +245% to about +178%. That is still a gain in the test. It is also a warning that the result was sensitive to a cost a beginner might round away.

On older data from 2020 to 2024, the same fee change was even harsher: about +77% became only about +8%. Same rules, older years, slightly higher cost. A result that shrinks from about +77% to about +8% is no longer the same story you would tell a friend. Part 4 stays with that older window. Here the point is only the fee.

Fewer trades, smaller damage

A quieter example was a slower 4-hour trend rule that traded much less often. It still dropped, but the damage was smaller, from about +26% to about +21% over 2 years. Fewer trades meant fewer bills.

That is the practical split. A rule that waits does not escape fees. It simply pays them less often, so the same fee rate hurts less. A rule that fires all day pays the round trip again and again. If each win is small, the 0.14% or 0.20% round trip can be most of the win.

Rules that entered very often

The opposite also showed up. Rules that entered very often, including some 15-minute entries under a higher-timeframe filter, could place around 30 or more trades a month. Several of those tests lost most of the account, even before the pattern itself was the argument. The activity was the problem. Each of those trades pays a fee to enter and again to exit, before you ask whether the picture on the chart was right.

One live-style Renko rule took more than 500 trades in 2 years. With real fills it lost about 72% at 0.07% fees, and about 86% at 0.10%. Part 1 already showed that real fills, not brick prices, were required before this rule could be believed. Once the fills were real, the higher fee made a bad result worse: about 72% lost became about 86% lost. More than 500 trades is a lot of chances for a small cost to compound against you.

What an honest fee test looks like

If a strategy needs many small wins to stay ahead, fees can erase the edge. An honest test should include realistic fees, then a slightly worse fee, and still make sense. If the plan only works when costs are almost zero, it is not ready for a live market.

You do not need a new indicator to do this check. Run the same rules at the base cost, here 0.07% per side, and again at the stricter cost, here 0.10% per side. If the story collapses, the edge was never larger than the bill. Part 3 looks at a different kind of missing piece: a stop, so one trade cannot sit open without a planned limit.

Key takeaways

  • You pay a fee to enter and again to exit, so costs add up faster than they look.
  • A strategy that trades often can go from winning to losing on fees alone.
  • Test with realistic fees, then a slightly higher fee, before you trust a result.

FAQ

What fee rates were used in these tests?

0.07% per side as a base, and 0.10% per side as a stricter check. A round trip is 0.14% at the base rate and 0.20% at the stricter rate, because you pay once to enter and once to exit.

How much did one pullback setup change when fees rose?

At 0.07% per side it was about +90% in 1 year and +245% in 2 years. At 0.10% per side those fell to about +69% and +178%. On 2020 to 2024 data, about +77% became about +8%.

Did a slower rule suffer as much?

Less. A 4-hour trend rule that traded less often dropped from about +26% to about +21% over 2 years. Fewer trades meant fewer bills.

What happened to a rule that traded hundreds of times?

One live-style Renko rule took more than 500 trades in 2 years. With real fills it lost about 72% at 0.07% fees and about 86% at 0.10%. Some 15-minute rules placed around 30 or more trades a month and lost most of the account.

This is education only and not financial advice. Trading involves risk, and you can lose money.